-
Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
New 28-bit DAC coming out.
- Thread starter GXAlan
- Start date
The first inside picture on this webpage clearly shows at least one ESS DAC, though the part number cannot be read: https://www.soundpure.com/p/millennia-imersiv-d-1-stereo-multi-path-dac-black/39443
IVX
Major Contributor
WOW, I didn't expect that. THD+N is worse than -114db, how did they implement it?
Other components are listed in the link:
www.audiosciencereview.com
New 28-bit DAC coming out.
The new D-1 DAC by Imersiv makes every DAC on the planet obsolete. I saw a clip in AudioXpress about the company winning an innovation prize from the SAE. You can learn more about it here: https://imersiv.com/ The SNR and distortion are 168dB below peak output which is +23dBu (10.9Vrms). No...
www.audiosciencereview.com
As there seems to be only one ESS ES9038PRO in an Immersiv D-1, which is a stereo device, that means that some of the 4 available on-chip internal DACs per channel are used for the lower path, and the remnant for the upper path.
So no advantage using any mic that needs phantom power, most codensers, and tube mics.
What does mic type have to do with that? The advantage of never running into gain setting issues is very real in practice with any mic.
Condensers mics that use phantom power (almost all of them) have built in preamps (thats what the phantoms for). These analog mic preamps have dynamic ranges much less than that preamp so whats the advantage of it? Ditto for tube mics. And if you think about the ambient noise in all recording spaces those ultra high S/N ratios are useless.
Gain staging is recording 101 and if your depending on your gear to do it for you you flunked.
EERecordist
Major Contributor
- Joined
- Feb 24, 2024
- Messages
- 1,200
- Likes
- 2,347
I hope we can keep ASR calm and friendly. Gain staging is a critical point of this topic. I have the opinion that the D1 is the start of higher dynamic range recording chains. Millennia Media, @signalpath makes microphone preamps, including with ADC, 24bit?, which hope someday are tested on ASR's Audio Precision, joining other microphone preamp tests. I hope Audio Precision and Rohde & Schwarz keep improving their test instruments. And I hope Analog Devices continues ADC innovation. Out of college I had a job offer from them, they are a great company, but I went to the dark side of digital.
I look at microphone specs often. From my understanding microphone self noise is measured on the same scale as maximum SPL. And I believe maximum SPL is commonly measured at 1% distortion. There may be some compression before the 1% is reached.
The classic U87 is speced at 117dB for 5% distortion and a self noise of about 12dB (current AI model.) We used them for close micing trombones. They have a switchable 10dB pad between the capsule and the impedance converter circuit inside the mic body. So the effective dynamic range is 105-115dB.
A classic close instrument microphone today is the DPA 4099. It has a max SPL of 138dB for 1% distortion against a noise floor of 23dB.
There have been digital microphones based on AES standards. Neumann's did not catch on. Schoeps is starting with its digital colette preamp. I have not looked lately, but Schoeps had optional switchable attenuators between the capsule and the preamp.
Generally the large diaphragm condensers have lower noise than the small diaphragm condensers. Austrian Audio and Rode have done good work reducing microphone self noise. And in recording and mixing polite gating (compression) can reduce noise in silent passages from a microphone. Polite gating is easily done in DAW mixing and by plugins.
Readers can study other microphones, capsule engineering, the SPL of acoustic instruments, and active circuit noise engineering to derive their own idea of what dynamic range is possible.
In the real world of recording it is hard to get below 25 dB(A) ambient. Right now in my quiet neighborhood home, I'm about 27dB(A) minimum on an afternoon. Concert halls have all kinds of human and HVAC noises. We sometimes recorded at night with an empty hall, but were still cursed by truck noise outside. A live concert with an audience for quiet classical music will be above 40 at the audience level. I regret not measuring our Hidley control rooms at the time. Typically the best control rooms are going to be 15-20dB(A) ambient.
That gets to what are the limits of the original source and in the end listening environment. With purely created digital music, the bit floor is the noise floor in the file.
Last edited:
?
The reason condenser microphones require power is because a bias voltage needs to be applied to the capsule.
Whether a microphone preamplifier is built-in or external is a separate issue from the type of microphone.
Obviously you have no idea what the conditions are in those large scale top tier installs that StageTec is dealing with. You may have many dozens of mic channels going into ADCs in a remote gear room/closet, not easily accessible for manual gain settings. You'd need racks full of mic preamps with remote control of the gain settings, and remote monitoring as well. Totally impractical. Good luck telling the producer the orchestra and singers have to do the whole opera one more time because you wrecked it, driving some mics into clipping or with too much headroom and noise, both of which may be only identified later in the production stage.
This is the reason why StageTec made the mic pre irrelevant by using an ADC that doesn't require attention to gain staging by balancing headroom vs. hiss.
Manual gain staging with conventional gear is only possible in small, localized setups like smaller recording studio, and when you have the time.
So before StageTec recording an orchestra was impossible?
Bias voltage is for electret mics, most studio condensers are not electret they use amplifiers in the mic body.
Last edited:
I think the Weiss OP2 opamp claims a 152dB SINAD, see https://weiss.ch/products/oem/op2-bp-2/
This would imply about 25 bits of dynamics, we are actually there. Since it is also marketed at the professional market, I would expect them to avoid absurd claims.
IVX
Major Contributor
I guess, THD+N -152db(the plot shows -153-155db) is a simulation-based number. The noise itself for 1nV/sqrt(Hz) and ideal zero impedances setup, makes -154db, H2 level -157db.
knownothing
Active Member
- Joined
- Apr 21, 2022
- Messages
- 175
- Likes
- 42
I looked at this link and am having some difficulty fully understanding the causal link between tmin and human hearing sensitivity and the fact that bit depth is the only factor contributing to the calculation of tmin. Be that as it may, if there is a direct relationship between tmin and human hearing sensitivity, based the equation presented in the link you provided low level (-40db), low frequency (20Hz) signals would have a tmin value of 11,000,000ns, or 11us, within the time resolution of human hearing. So at the very least, wouldn’t this dictate that for very high fidelity reproduction that 24bits or even 20 bits would be an improvement?
RandomEar
Major Contributor
- Joined
- Feb 14, 2022
- Messages
- 1,145
- Likes
- 2,953
My understanding is as follows: As long as your sampling rate is high enough to have a Nyquist frequency above the upper threshold of human hearing (commonly quoted as 20 kHz), you can reproduce all relevant frequencies faithfully. This is the case for common audio formats, which start at 44.1 kHz sampling rate and go up to a couple hundred kHz. At that point, only your bit depth detemines the loss of precision in your recording. This loss is called quantization error. In itself it is an amplitude error, but of course a slight difference in amplitude at the sample points would also shift a sine wave of a fixed frequency slightly to the left or right in the reconstructed signal. So in the end, this amplitude error due to quantization also limits the time resolution of your reconstruction. Other members with a deeper understanding of the topic are welcome to correct me on this, in case this explanation isn't quite accurate.
After checking, it seems that it's not totally clear which auditory threshold would apply here. There are a multitude of results, some binaural, some monaural, for different stimuli like double clicks or time differences between the ears. The lowest number I have seen for any stimulus was around 10 µs, which would indeed be exactly the magnitude you mention.
I guess it depends: Due to the equal loudness contour, our ears are much less sensistive at those frequencies. The difference is easily 30 dB compared to the 1 kHz baseline, so a signal at -40 dB would effectively sound like -70 dB, which would likely just get masked by everything else happening in the music. And at higher amplitudes, you would gain time resolution again. It's also not totally clear if a time difference of 10 µs would be less or more audible at 20 Hz than at the frequencies where the relevant experiment was conducted (I think it was 300 and 3600 Hz). Overall, this looks like a borderline result to me.
So to be absolutely sure, I guess you could either say "go for 20 bits and forget about it", or you could repeat one of those experiments under the critical conditions you listed.
popej
Addicted to Fun and Learning
- Joined
- Jul 13, 2023
- Messages
- 775
- Likes
- 682
There is no loss of precision for signals below Nyqust frequency. Please note, that if we talk about signal with frequency, we mean periodic wave, which consist of multiple samples. There is no shift of signal caused by error of a single sample. Actually with a proper dithering, quantization error is turned into noise. And you can hear periodic signals below noise level.
The actual DA convertor output is stepped, i.e. it contains above-nyquist ultrasonic frequencies which must be removed (the steps "smoothed out") by the analog reconstruction filter. For 44.1 the nyquist is 22kHz - it's very hard to make an analog filter so steep to keep 20kHz and "completely" eliminate 22kHz. So e.g. upsampling 8x moves the nyquist to 176kHz - the analog reconstruction filter now must filter out everything above 176kHz, while keeping 20kHz - much easier to do.
Oversampling on the ADC side is done for the very same reason - the input analog filter before the ADC (again "completely" eliminating from nyquist up) does not need to be so steep because the AD conversion nyquist is moved much higher by the oversampling.
Similar threads
